The takeaway
Apple is turning Siri into an agent runtime with App Intents as the action schema, which makes a machine-readable action surface the price of admission for any app or service that wants to be operated by an assistant.
Why it matters for builders
If your product or internal tool can be described as actions, expose them as intents now. Agents that can call a declared capability are reliable and cheap; agents that must drive a UI are brittle and expensive. Expect Apple to make third-party app actions the default integration point for Siri, the same way Shortcuts became the automation layer for iOS. Teams shipping automation should model their workflows as named, permission-scoped actions with typed inputs so both a human assistant and a headless agent can invoke them.
Apple is racing to turn Siri into a genuine AI agent
Apple shipped a working, personalized Siri AI inside iOS 27 and then discovered the harder problem: an assistant that answers requests is not the same thing as an agent that finishes jobs. According to Bloomberg's Mark Gurman, CEO John Ternus held a closed-door meeting with software chief Craig Federighi in recent weeks and pushed him to accelerate the roadmap and move beyond incremental Siri fixes.
What Bloomberg reported
Gurman's Power On newsletter describes a company that knows it is behind. OpenAI's Dots, Meta's Muse and xAI's Grok already behave as persistent agents: they hold a goal, decide what to do next and keep working without the user restarting them at every step. Siri in iOS 27.2 and the iOS 27.3 beta still operates request by request.
Ternus is not hiding the gap. Speaking about Siri AI, he described it as "primarily request-based today" but built on an "orchestrator-driven architecture" that leaves "room to expand capabilities over time" according to 9to5Mac.
A two-step roadmap
The first step is already in developer beta. iOS 27.3 gives Siri fine-grained control over Apple's own apps - rotating and cropping photos, sorting Safari tabs, scheduling messages, building calendar events with invitees, managing files. Gurman expects those capabilities to reach users before long.

The second step is the one that matters. Third-party app control is expected in the spring, letting Siri place an order through a delivery app or book a table through a reservation app. Only after that does Apple intend to make Siri genuinely agentic - holding objectives across sessions rather than executing single commands - possibly as early as 2027 (The Verge).
The integration layer is already shipping
The technical groundwork is less dramatic than the roadmap talk but more consequential. In the iOS 27.3 beta, Apple opened an early developer preview that lets any app expose its content and actions to Siri through App Intents. Apple's own example: a user asks Siri to send their favorite sneakers to a partner, and Siri finds the items, resolves the contact and completes the handoff across apps (9to5Mac).
That is a declarative contract, not screen scraping. Apple is asking developers to describe what their app can do, and letting the assistant decide when to call it.
Why it matters
Every assistant that becomes an agent eventually needs a machine-readable action surface. Apple is building that surface now, with App Intents as the schema, Shortcuts as the fallback and Siri as the runtime. The companies that publish clean intents early will be the ones an agent can actually operate; the ones that do not will be reachable only through fragile computer-use automation.
For teams building automation products, the read is straightforward: the mobile assistant is turning into an execution environment. Designing for an API contract now is cheaper than retrofitting one when Siri starts pursuing goals on its own.
The Automation Brief
Read 5 AI stories instead of 50.
The essential moves in AI agents, models, automation and infrastructure — filtered for builders and operators, with the part that actually matters.
No noise. Unsubscribe anytime.
Editorial notes
Stefan Trbojevic
n8n Lab Editorial
9 November 2026
9 November 2026
Sources
AI disclosure: AI assisted with research and drafting. Factual claims are reviewed by an editor.




